## wpiea2026015-source-pdf - 0.03 percent in consumption-equivalent terms. The key mechanism is that workers are better

## Source details

**Canonical URL:** [wpiea2026015-source-pdf - 0.03 percent in consumption-equivalent terms. The key mechanism is that workers are better](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026015-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026015-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026015-source-pdf.pdf.json)

---

### Literature review and paper contributions
- Empirical strands on job-retention schemes (STW): cross-country macro evidence and within-country/administrative variation.
- Cross-country patterns: STW take-up highest among firms hit by negative demand shocks; more prevalent where firing costs are high, firm-specific human capital is valuable, and downward nominal wage rigidity is strong.
- Selected country findings:
  - Germany: STW helped limit job destruction during the GFC.
  - Evidence that STW stabilizes employment but may induce inefficient reductions in hours and affect outsiders.
  - Supported firms reduce hours per worker and increase employment relative to untreated firms.
  - Heterogeneity and targeting/design of schemes matter.
- Theoretical literature: stabilization benefits emphasized; many models omit occupation switching and costs of occupational reallocation.
- This paper’s contributions:
  - Worker-level cross-country evidence: more generous job-retention schemes associated with lower separation probabilities during downturns, especially for low-wage workers.
  - A search-and-matching model with endogenous occupational choice, costly occupational switching, and downward nominal wage rigidity.
  - Calibration to the pre-crisis United Kingdom to quantify distributional and fiscal implications.

### Data, stylized facts, and descriptive statistics
- Data source and sample:
  - European Union Statistics on Income and Living Conditions (EU-SILC) for 12 EU/OECD countries (Austria, Belgium, Denmark, Estonia, Finland, Italy, Netherlands, Portugal, Slovenia, Spain, Sweden, United Kingdom) covering 2003–2018.
  - Sample restricted to ages 20–65; over 1.8 million individuals (Total 1,879,861; Avg Annual Obs 129,639).
  - Occupational categories: one-digit ISCO-08 codes (nine occupation categories).
- Stylized facts:
  - Earnings loss after one year unemployment (EUE): average change −41.6 log points; stayers −32.5 log points; occupation switchers additional −21.5 log points penalty.
  - On-the-job occupation switchers (EE) obtain additional earnings: EE average gain 6.5 log points; EE stayers 6.1 log points; EE on-the-job switchers additional 2.3 log points.
  - Job separation negatively correlated with prior-period wages: ln(Wage) coefficients on Job Separation −0.0253 ***, −0.0239 *** with country fixed effects.
  - Low-wage workers have higher separation rates and are disproportionately affected in recessions.
  - Countries with high job retention policy spending above the median show lower job separation probabilities during recessions (job separation probability is 1.2 percentage points lower in high-retention countries, based on column (5) interpretation).

### Empirical specification and key regression results
- Separation regression:
  - Separation_ict = α_c + β_t + γ Rec_ct + δ HighReten_c + θ Rec_ct × HighReten_c + λ X_ict + e_it
  - HighReten_c = 1 if country’s average retention policy is above the median.
  - Rec_ct = 1 if country c is in recession in year t (Harding and Pagan (2002)).
  - X_ict includes individual controls and country time-varying controls (output gap, ln GDP per capita, UI and reallocation policy controls).
- Selected coefficients (Table 3):
  - Rec: 0.0200 ***, 0.0158 ***, 0.0108 ***, 0.00610 **, 0.00570 ** across specifications.
  - HighReten (col 1): −0.0129 ***.
  - Rec × HighReten: −0.0228 ***, −0.0214 ***, −0.0192 ***, −0.0122 ***, −0.0122 ***.
  - Observations: 188,700 (cols 1–3); 180,315 (col 4); 178,809 (col 5).
  - R-squared examples: 0.00374; 0.00807; 0.00926; 0.0104; 0.0184.
- Differential impacts for low-wage workers (triple-interaction; Table 4, col (5)):
  - ln(Wage) × Rec × HighReten = 0.0240 ∗∗∗ (standard error 0.00835).
  - Rec = 0.0536 ∗∗∗ (0.0151).
  - ln(Wage) = −0.0301 ∗∗∗ (0.00258).
  - ln(Wage) × Rec = −0.0209 ∗∗∗ (0.00663).
  - Rec × HighReten = −0.0687 ∗∗∗ (0.0214).
  - ln(Wage) × HighReten = 0.00460 (0.00406).
  - Observations = 150,288; R-squared = 0.0292.
- Illustrative interpretation using wage ln-percentiles 25th = 2.03 and 75th = 3.04:
  - Low-wage worker has 3 percentage point higher chance of losing job in normal times compared to high-wage worker (ln(Wage) effect).
  - During a recession, absent retention policy, low-wage worker’s chance increases by 2.1 percentage points (ln(Wage) × Rec).
  - In a country with strong retention policy, this additional increase is fully mitigated by 2.4 percentage points (ln(Wage) × Rec × HighReten).

### Theoretical model — setup and main propositions
- Equilibrium objects:
  - Policy functions {V_k(ε,z), W_k(ε,z), ˆk_k(ε,z)} given labor market tightness {θ_k(ε,h,z)}, wages {w_k(ε,h)}, tax schedule {τ(z)}.
  - Firms’ policy functions {G_k(ε,z), J_k(ε,h,z), r_k(ε,h,z), s(ε,h,z)} imply free-entry tightness schedule satisfying condition (6).
  - Population density distributions {ˆx_k(h,z)} satisfy (9) and (10) and stationarity ˆx_j(h,z) = x_j(h,z).
- Shock and retention mechanics:
  - Occupation-specific shock: z_k = (1−φ_k)̄z_k (equation (14)); φ_k ∈ [0, (1−γ)(1−T_u / (̄z_k H_0))] (equation (15)).
  - Profit cutoff without retention: ε_k(h,φ_k) ≡ ( 1 / (1−φ_k)^{1−γ} ) T_u / (z_k h) (equation (16)).
  - With retention transfer cap T_r > 0, new cutoff ε_k^r(h,φ_k) ≡ ( 1 / (1−φ_k)^{1−γ} ) (T_u − T_r) / ((1−γ) z_k h) (equation (17)).
  - Measure of saved matches: ε_k^1(h,φ_k) − ε_k^r(h,φ_k) = T_r ((1−γ) − φ_k) z_k h^{-1} (equation (18)); increasing in φ_k and decreasing in h.
- Lemma and propositions:
  - Lemma 1: retention saves more matches the deeper the recession and is more effective for low-productive firms.
  - Proposition 1: retention transfers save more matches when recessions are deeper, save more low-productive matches if human capital distribution non-increasing, and save more matches in occupations more severely impacted.
  - Proposition 2: retention transfers reduce the public deficit unequivocally if T_r ≤ 2 T_u (under stated assumptions).

### Calibration targets and parameter choices (UK benchmark)
- Two occupations: K = {R, N} (Routine R and Non-routine N).
- Aggregate AR(1) parameters:
  - ρ_R = 0.42, σ_R = 0.015; ρ_N = 0.37, σ_N = 0.008.
- Preferences and matching:
  - CES utility exponent α = 2.
  - Matching function parameter ν = 1.6.
  - Bargaining power γ = 0.70.
- Selected calibrated parameter values (Table 5 excerpts):
  - T_r = 0.000 (Baseline)
  - z_R = 1.000
  - r = 0.010
  - α = 2.000
  - β = 0.960
  - γ = 0.700
  - ν = 1.600
  - ρ_R = 0.422
  - σ_R = 0.015
  - ρ_N = 0.366
  - σ_N = 0.008
  - z_N = 0.955
  - T_u = 0.155
  - δ = 0.010
  - χ = 0.001
  - α_Bh = 7.255
  - β_Bh = 4.855
  - σ_ε = 0.355
- Calibration fit (selected moments, Table 6):
  - Routine employment share: Data 0.553, Model 0.547.
  - Unemployment benefits-GDP ratio: Data 0.002, Model 0.002.
  - Unemployment rate before GFC: Data 0.051, Model 0.052.
  - Unemployment rate during GFC: Data 0.072, Model 0.069.
  - Labor income normalized percentiles (data): 0.01/0.04/0.09/0.15/0.22; (model): 0.08/0.14/0.21/0.27/0.34.
  - Volatility normalized labor income: Data 0.874, Model 0.779.

### Quantitative experiments and baseline results (transitory shock)
- Shock design:
  - Asymmetric negative productivity shock replicating UK GFC/European Debt Crisis experience.
  - Routine employment contraction: 6.9 percentage points; non-routine: 2.4 percentage points.
  - Shock lasts one period; transition back to steady state lasts 10 periods (10 periods represent 2 years and 6 months).
- Baseline (no retention):
  - Unemployment rate increases by 1.6 percentage points after the shock.
  - Share of workers in routine occupations falls by almost two percentage points.
  - Employment in non-routine occupations increases by less than one percentage point.
  - Worker reallocation: high-human-capital workers move toward non-routine occupations.
- Retention policy (designed to retain half of otherwise lost matches):
  - Unemployment rises by 0.8 percentage points with retention (instead of 1.6 percentage points baseline).
  - Employment routine falls by 1.5 percentage points with policy (instead of 2 percentage points baseline).
  - Employment non-routine rises by 0.8 percentage points with policy (instead of 0.5 percentage points baseline).
  - Policy effects: reduces spike in unemployment, speeds transition back to steady state, reduces occupation switches from routine to non-routine.

### Fiscal impacts and budget dynamics (transitory shock)
- Main fiscal drivers: unemployment benefits, labor tax revenues, retention transfers.
- Government spending and revenue outcomes (with retention vs baseline):
  - Unemployment insurance spending decreases from 2.7% of GDP (baseline) to 2.4% of GDP (with retention): decrease of 0.3% of GDP.
  - Retention policy direct fiscal cost is almost 0.02% of GDP compared to the 0.3% savings in UI.
  - Baseline fiscal deficit: 0.6 percentage points during shock; return to balanced budget in about seven periods.
  - With retention: fiscal deficit 0.3 percentage points during shock; return to balanced budget in five periods.
  - Interpretation: retention policy halves the fiscal deficit during the shock period and shortens time to fiscal balance.

### Welfare, distributional effects, and transfer incidence
- Aggregate welfare:
  - Retention policy yields positive welfare gains of about 0.06% increase in permanent consumption-equivalent (baseline transitory-shock experiment).
- Distributional outcomes:
  - Three lowest human capital deciles receive about 50% of retention transfers.
  - Routine occupation accrues 61% of all transfers.
  - Lowest deciles’ welfare increases by almost 0.15%; higher human capital workers benefit less.
  - Net effect: all workers better off with retention policy because lower government deficit implies lower transfer burdens.
- Policy interpretation: welfare gains concentrated among low-income workers who benefit most from preserved matches and avoided costly reallocation.

### Robustness: longer transitory shocks, permanent shocks, and bureaucratic costs
- Two-period transitory shock scenarios:
  - No retention: unemployment rises from 7.3 percent in period 0 to 8.2 percent in period 1.
  - Retention only in first period: unemployment 6.4 percent in period 0, rising to 7.7 percent in period 1 after policy removal.
  - Retention in both shock periods: unemployment limited to 6.3 percent (period 0) and 6.8 percent (period 1), stabilizing employment.
  - Fiscal effects (relative to no retention):
    - Retention only first period: deficit declines by 0.31 percent of GDP in period 0; improvement weakens to 0.19 percent reduction in period 1 once policy removed.
    - Retention both periods: deficit drops by 0.36 percent in period 0 and further reduces by 0.53 percent in period 1 (illustrating prevention of separations throughout the shock).
  - Welfare gains (permanent consumption-equivalent):
    - Retention only first period: ≈ 0.07 percent.
    - Retention throughout both shock periods: ≈ 0.14 percent.
    - Doubling of lifetime welfare benefits from single-period to full-duration coverage consistent across deciles.
- Permanent shock (2.4 percentage points permanent productivity decline):
  - Without retention, unemployment adjusts almost immediately to steady state 6 percent.
  - One-period retention delays separations but unemployment converges to same steady state 6 percent once support withdrawn.
  - Fiscal improvement from one-period retention is "on the order of one fifth of a percentage point of GDP."
  - Welfare gain modest: around 0.03%.
  - Mechanism: retention supports on-the-job search and bargaining power even when preserved matches are not viable long-run.
- Bureaucratic cost sensitivity:
  - Adding bureaucratic cost calibrated at 0.5% and 1% of GDP.
  - For every £1 spent on retention transfers, government spends £31 or £63 on bureaucratic costs respectively (for 0.5% and 1% bureaucratic cost calibrations).
  - Baseline policy cost (no bureaucratic cost) is 0.02% of GDP.
  - Large bureaucratic costs financed via lump-sum transfers increase government deficit, reduce consumption, and lower welfare for all agents.
  - With 1% of GDP bureaucratic cost, retention policy can cease to be a Pareto improvement; high-human-capital workers can experience welfare losses.

### Key quantitative takeaways and policy implications
- A retention scheme calibrated to offset roughly half of the observed rise in unemployment during calibrated historical episodes would have materially dampened labor-market deterioration.
- Aggregate welfare rise: 0.06 percent in consumption-equivalent terms in the baseline transitory-shock experiment; gains concentrated among low-income workers.
- Fiscal performance: retention can be deficit-reducing during transitory shocks by lowering UI outlays and preserving tax revenues; in baseline experiment retention halves the fiscal deficit during the shock period.
- Policy design matters:
  - Targeting and duration are critical—covering the full duration of a persistent shock materially strengthens welfare and fiscal outcomes.
  - Implementation inefficiencies and bureaucratic costs can negate fiscal and welfare gains; low administrative costs are key for Pareto improvements.

_Italic: Source: Content unit "wpiea2026015-source-pdf - 0.03 percent in consumption-equivalent terms. The key mechanism is that workers are better" (excerpts provided)._

### 0.03 percent in consumption-equivalent terms. The key mechanism is that workers are better

### wpiea2026015-source-pdf - 0.03 percent in consumption-equivalent terms. The key mechanism is that workers are better

### Literature Review
- Empirical literature on job-retention schemes (short-time work, STW) grouped into two strands: cross-country macro evidence and within-country/administrative variation.
- Cross-country evidence: STW take-up highest among firms hit by negative demand shocks; more prevalent where firing costs are high, firm-specific human capital is valuable, and downward nominal wage rigidity is strong (Lydon et al. (2019)).
- Country-specific studies (examples and findings):
  - Germany: Boeri and Bruecker (2011) find STW helped limit job destruction during the GFC.
  - Cahuc and Carcillo (2011): STW stabilizes employment but may induce inefficient reductions in hours and adversely affect outsiders.
  - Balleer et al. (2016): STW stabilizes employment more in rigid than flexible labor markets.
  - Giupponi and Landais (2023): Supported firms reduce hours per worker and increase employment relative to untreated firms.
  - Cahuc et al. (2018, 2021): STW saved jobs among firms with large revenue declines; substantial heterogeneity and importance of targeting and scheme design.
- Theoretical literature highlights stabilization benefits but often omits key margins:
  - Tilly and Niedermayer (2017): modest welfare gains in life-cycle search model; no occupation switching.
  - Balleer et al. (2016): downward nominal wage rigidity and firing costs; no heterogeneous recession impacts across occupations or costs of occupational reallocation.
  - Dengler and Gehrke (2021): DSGE with endogenous separations and policy rule; no occupation-switching channel.
  - Birinci et al. (2021): payroll subsidies and UI during a pandemic; limited but potentially welfare-improving role for retention-type policies.
- Contributions of this paper:
  - Worker-level cross-country evidence: more generous job-retention schemes associated with lower separation probabilities during downturns, especially for low-wage workers.
  - Development of a search-and-matching model with endogenous occupational choice, costly occupational switching, and downward nominal wage rigidity.
  - Calibration to the pre-crisis United Kingdom to quantify distributional and fiscal implications: welfare gains concentrated among low-income workers; saved matches reduce UI outlays and increase tax revenues; a well-designed scheme can be deficit-reducing despite the wage subsidy.

### Stylized Facts — Overview
- Data source: European Union Statistics on Income and Living Conditions (EU-SILC) for 12 EU/OECD countries from 2003 to 2018.
- Sample restricted to individuals aged between 20 and 65 years.
- Sample size: over 1.8 million individuals across 12 countries (Austria, Belgium, Denmark, Estonia, Finland, Italy, Netherlands, Portugal, Slovenia, Spain, Sweden, and United Kingdom).
- Occupational categories: one-digit ISCO-08 codes (nine occupation categories).
- EU-SILC limitations: no industry information; captures only large occupational switches due to one-digit ISCO-08 categories; panel up to four years; data cover 2003–2018.
- Advantage: cross-country heterogeneity enables exploration of retention scheme use and occupational switching.

### Cost of Job Loss and Occupational Switch (Table 1 findings)
- Workers reemployed after one year of unemployment ("EUE") have average earnings change of -41.6 log points (column 1 labeled "EUE").
- Reemployed stayers (returned to same occupation) lost -32.5 log points.
- Reemployed occupation switchers experienced an earnings loss of -54 log points (=-32.5-21.5), implying an additional switching penalty of -21.5 log points relative to stayers.
- The additional earnings penalty from switching upon reemployment is robust to country year fixed effects, year fixed effects, macro controls (output gap and ln GDP per capita), and individual characteristics (gender, age, age-squared, high-skill dummy) (column 3).
- Workers who remained employed ("EE") had an average earnings gain of 6.5 log points (column 4).
  - EE stayers (same occupation) experienced gains of 6.1 log points.
  - EE on-the-job switchers obtained an additional 2.3 log points in earnings (column 5).
- Interpretation: on-the-job occupational switching is associated with voluntary additional earnings gains, while occupational switching upon reemployment after unemployment is involuntary and costly.

- Table 1 (selected reported statistics):
  - Occ. Switching coefficient in EUE sample: -0.215 ***
  - Occ. Switching coefficient in EUE with controls: -0.206 ***
  - Occ. Switching coefficient in EE sample: 0.0226 ***
  - Occ. Switching coefficient in EE with controls: 0.0130 **
  - Constant (EUE, column 1): -0.416 ***
  - Constant (EUE stayers, column 2): -0.325 ***
  - Constant (EE, column 4): 0.0648 ***
  - Observations across columns: 2,126; 2,126; 2,109; 172,801; 172,801; 171,808
  - R-squared values reported (selected): 0.00938 (col 3), 0.0304 (col 6)
- Notes: Standard errors clustered at country and year level; significance levels: * p<0.10, ** p<0.05, *** p<0.01. Macro controls include output gap and ln GDP per capita. Individual characteristics include age, age-squared, female dummy, marital status dummy, and high skill dummy (college and above degree).

### Job Separation (Table 2 findings)
- Negative correlation between job separation and prior-period wages: lower past wages associated with higher separation probability.
- Table 2 coefficients (selected):
  - ln(Wage) effect on Job Separation: -0.0253 ***
  - ln(Wage) effect with country fixed effects: -0.0239 ***
  - Rec dummy coefficient (column 1): 0.0311 *
  - Interaction Rec x ln(Wage): -0.0101 * (column 3); -0.00796 ** (column 4)
  - Constant examples: 0.0947 *** (col 1), 0.0713 (col 2)
  - Observations: 188,365; 187,097; 157,271; 156,157
  - R-squared values: 0.0164; 0.0264; 0.0179; 0.0278
- Interpretation: lower-wage workers have higher job separation rates, and low-wage workers experience even higher separation rates during recessions.
- Notes: Dependent variable is dummy for job in year t-1 and unemployed in year t. Rec defined following Harding and Pagan (2002). Standard errors clustered at country and year level. Macro controls include output gap and ln GDP per capita. Individual characteristics include age, age-squared, female dummy, marital status dummy, and high skill dummy (college and above degree).

### Role of Retention Policies in Mitigating Job Separation (Table 3 findings and regression setup)
- Regression specification:
  - Separation_ict = α_c + β_t + γ Rec_ct + δ HighReten_c + θ Rec_ct × HighReten_c + λ X_ict + e_it
  - HighReten_c = 1 if country’s average retention policy is above the median.
  - Rec_ct = 1 if country c is in recession in year t (Harding and Pagan (2002) dating: recessions defined as contiguous blocks of years with negative real GDP growth).
  - X_ict includes individual controls (age, age-squared, gender, marital status, high skill) and country time-varying controls (output gaps, ln GDP per capita, policy variables including unemployment insurance and reallocation policies).
  - Definition of broad retention policies uses OECD Labor Market Policies database categories: benefits administration (12), workplace training (22), apprenticeship (24), employment incentives for maintenance (keeping jobs) (42), partial unemployment benefits (82), part-time unemployment benefits (83).
  - Reallocation policies defined as sum of OECD items: placement administration (11), institutional training (21), integrated training (23), employment incentives for recruitment (hiring) (41), direct job creation (60), start-up incentives (70), early retirement.
- Key regression results (Table 3, selected):
  - Rec coefficient: 0.0200 *** (col 1), 0.0158 *** (col 2), 0.0108 *** (col 3), 0.00610 ** (col 4), 0.00570 ** (col 5)
  - HighReten coefficient (col 1): -0.0129 ***
  - Rec x HighReten interaction: -0.0228 *** (col 1); -0.0214 *** (col 2); -0.0192 *** (col 3); -0.0122 *** (col 4); -0.0122 *** (col 5)
  - Constant examples: 0.0406 *** (col 1), 0.0351 *** (col 2), 0.0361 *** (col 3), 1.281 ** (col 4), 1.445 ***
  - Observations: 188,700 (cols 1–3), 180,315 (col 4), 178,809 (col 5)
  - R-squared values: 0.00374; 0.00807; 0.00926; 0.0104; 0.0184
- Interpretation:
  - Countries with high job retention policy spending above the median have lower job separation probabilities during recessions.
  - Based on column (5), the job separation probability is 1.2 percentage points lower in countries with job retention policy spending above the median.
- Notes: Dependent variable is dummy for job in t-1 and unemployed in t. Standard errors clustered at country and year level. Macro controls include output gap and ln GDP per capita. Individual characteristics include age, age-squared, female dummy, marital status dummy, and high skill dummy (college and above degree).

*Source: Content unit "wpiea2026015-source-pdf - 0.03 percent in consumption-equivalent terms. The key mechanism is that workers are better" (excerpts provided).*

### 3.5    Differential Impacts of Retention Policies for Low-wage Workers

### 3.5    Differential Impacts of Retention Policies for Low-wage Workers

### Empirical extension: triple interaction specification
- The baseline specification (equation (1)) is extended to include triple interactions among ln(Wage), Rec (recession dummy), and HighReten (high retention policy dummy), together with their respective double interactions.
- The main variable of interest is ln(Wage)xRecxHighReten.

### Main empirical results (Table 4, column (5) highlights)
- Triple interaction:
  - ln(Wage)xRecxHighReten = 0.0240 ∗∗∗ (standard error 0.00835)
- Double and single interaction and main effect coefficients (column (5)):
  - Rec = 0.0536 ∗∗∗ (0.0151)
  - ln(Wage) = -0.0301 ∗∗∗ (0.00258)
  - ln(Wage)xRec = -0.0209 ∗∗∗ (0.00663)
  - RecxHighReten = -0.0687 ∗∗∗ (0.0214)
  - ln(Wage)xHighReten = 0.00460 (0.00406)
- Model fit and sample:
  - Observations = 150,288
  - R-squared = 0.0292
- Controls included in column (5): Country FE, Year FE, Macro Controls, Policy Controls, Individual Characteristics.

### Illustrative interpretation (quantitative example from text)
- Wage percentiles: the 25th percentile and 75th percentile of ln(wage) are 2.03 and 3.04, respectively.
- Based on column (5) coefficients:
  - A low-wage worker has a 3 percentage point higher chance of losing their job in general compared to a high-wage worker (ln(Wage) during normal, non-recession times).
  - During a recession, absent a retention policy, the low-wage worker’s chance of losing their job increases by 2.1 percentage points (ln(Wage)xRec).
  - If the low-wage worker lives in a country with a strong retention policy, this additional increase is fully mitigated by 2.4 percentage points (ln(Wage)xRecxHighReten).

### Stylized facts established
- First: The earnings loss for those who went through an unemployment spell is significant and is amplified for those who had to switch occupations.
- Second: Job separation rates for low-wage workers are higher than those of their counterparts and are particularly high during recessions.
- Third: Job separation rates are overall higher during recessions but are mitigated in countries with strong job retention schemes.
- Fourth: The mitigating effect on the job separation rate during recessions seems to be larger for low-wage workers in countries with high retention policies.

### Data and inference notes
- Standard errors are clustered at country and year level. Significance levels: ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01.
- HighReten is defined as a dummy for which the average retention policies are above the 50th percentile in the sample.
- Policy controls include unemployment insurance spending and reallocation policies spending as percent of GDP. Macro controls include output gap and ln GDP per capita. Individual characteristics include age, age-squared, female dummy, and high-skill dummy (college and plus education).
- Limitation noted: examining the impact of retention policies in mitigating earnings loss (as studied in Table 1) is not possible due to the small sample size, particularly for those who get reemployed after one year of unemployment spell.

### Link to model motivation
- Motivated by these stylized facts, the next section builds a search-and-matching model with endogenous occupational choices that incorporates job retention policies and the heterogeneity and risks across occupations.

*Source: 3.5 Differential Impacts of Retention Policies for Low-wage Workers (from the supplied PDF content).*

### 1.  Given the labor market tightness

### 1.  Given the labor market tightness

### Equilibrium system and policy functions
- Agents’ policy functions {V_k(ε,z), W_k(ε,z), ˆk_k(ε,z)} solve equations (2), (3), and (5) given:
  - labor market tightness schedule {θ_k(ε,h,z)},
  - wages schedule {w_k(ε,h)},
  - tax schedule {τ(z)}.
- Firms’ policy functions {G_k(ε,z), J_k(ε,h,z), r_k(ε,h,z), s(ε,h,z)} imply the labor market tightness schedule {θ_k(ε,z)} satisfies the free-entry condition (6).
- Labor income tax schedule {τ(z)} balances the government budget (13) given:
  - population density distributions {ˆx_k(h,z)},
  - wages schedule {w_k(ε,z)}.
- Population density distributions {ˆx_k(h,z)} satisfy (9) and (10) given the occupation policy function {ˆk_k(ε,h,z)}.
- Stationarity condition for distribution across occupations j ∈ K ∪ {0}:
  - ˆx_j(h,z) = x_j(h,z).

### Theoretical analysis — setup and key assumptions
- Occupation-specific disturbance modeled as an unanticipated multiplicative shock to average match productivity in occupation k:
  - pre-shock parameter ̄z_k; post-shock z_k = (1−φ_k)̄z_k (equation (14)).
  - φ_k indexes severity; φ_k = 0 corresponds to no disruption.
- Support restriction for φ_k to ensure non-negative post-shock productivity and interior retention decisions:
  - φ_k ∈ [0, (1−γ)(1−T_u / (̄z_k H_0))] (equation (15)).
- Partial equilibrium abstraction: distribution of matches and inflow of new matches treated as fixed when shock hits.
- Assumption 1:
  - (i) Worker human capital bounded below by H_0 > 0.
  - (ii) Occupation-specific productivity z_k = (1−φ_k)̄z_k with φ_k ∈ [0, (1−γ)(1−T_u/(̄z_k H_0))].
  - (iii) Density of matches over occupation-specific human capital ε_k is uniform.

### Firms’ profits, cutoffs, and retention transfers
- Firms’ profit under static Nash-bargaining:
  - I_k(ε,h,z) ≡ z_k ε_k h − w_k(ε,h) = (1−γ)((1−φ_k)^{1−γ} z_k ε_k h − T_u).
- Profit cutoff without retention:
  - ε_k(h,φ_k) ≡ ( 1 / (1−φ_k)^{1−γ} ) T_u / (z_k h) (equation (16)).
  - Matches with ε_k below ε_k receive negative profits; above receive positive profits.
- Retention policy: transfers bounded by T_r > 0; firms receive transfers up to covering wage bill to that upper limit.
- Firms’ profits under retention policy:
  - I_k^r(ε,h,z) =
    - z_k ε_k h − w_k(ε,h) if z_k ε_k h − w_k(ε,h) ≥ 0;
    - 0 if z_k ε_k h − w_k(ε,h) < 0 and z_k ε_k h − w_k(ε,h) + T_r ≥ 0;
    - z_k ε_k h − w_k(ε,h) if z_k ε_k h − w_k(ε,h) + T_r < 0.
- New cutoff under retention:
  - ε_k^r(h,φ_k) ≡ ( 1 / (1−φ_k)^{1−γ} ) (T_u − T_r) / ((1−γ) z_k h) (equation (17)).
- Measure of saved matches (uniform ε_k distribution):
  - ε_k^1(h,φ_k) − ε_k^r(h,φ_k) = T_r ((1−γ) − φ_k) z_k h^{-1}. (equation (18))
  - This is increasing in φ_k (deeper recessions save more matches for fixed T_r) and decreasing in h (saves more low-human-capital matches).

### Lemma and Propositions — main theoretical results
- Lemma 1 (Effectiveness across productivity):
  - Under Assumption 1 and fixed maximum retention transfer, a retention policy saves more firms’ matches the deeper the recession within an occupation k and human capital h.
  - Retention policy is more effective in safeguarding matches of low-productive firms within a given h and occupation k.
- Proposition 1 (Retention transfers effectiveness) under Assumptions 1 and fixed T_r > 0:
  - i) A retention scheme saves more matches the deeper the recession (defined via φ_k comparisons).
  - ii) If the human capital distribution is non-increasing, the retention scheme saves more low-productive matches.
  - iii) If average productivity is the same across occupations (̄z_k = ̄z), the retention scheme saves more matches in occupations more severely impacted by the recession.
- Proposition 2 (Retention transfers efficiency) under Assumptions 1 and a recession φ > 0 such that each occupation k contraction is below (1−γ)(1−T_u/(z_k H_0)):
  - A retention transfer scheme reduces the public deficit unequivocally if T_r ≤ 2 T_u.

### Quantitative analysis — calibration target and scope
- Calibration focus: U.K. economy around the Global Financial Crisis (GFC) and European debt crisis.
- Two occupations only: K = {R, N} (Routine R and Non-routine N).
- Aggregate productivity AR(1) process (occupation-specific):
  - ln(z_k′) = (1−ρ_k) ln(z_k) + ρ_k ln(z_k) + σ_k ε, with ε ∼ N(0, 1).
  - Estimated parameters:
    - ρ_R = 0.42, σ_R = 0.015 (routine: more persistent and more volatile).
    - ρ_N = 0.37, σ_N = 0.008.
- Agents’ preferences:
  - CES utility u(c) = c^{1−α} / (1−α).
  - α = 2 (intertemporal elasticity parameter).
- Matching function M(s,v) = s v (s^ν + v^ν)^{1/ν} with ν = 1.6.
- Bargaining power: γ = 0.70.
- Agents’ productivity decomposition:
  - e ∈ R^K_+ decomposed into human capital h = ||e||_2 and occupational advantage ε = e / ||e||_2.
  - Human capital h ∼ Beta(α_Bh, β_Bh); calibrated to match mean and standard deviation of labor income distribution from EU-SILC (UK, 2005–2016).
  - Empirical normalized labor income percentiles: 10th/25th/50th/75th/90th = 0.11, 0.36, 0.82, 1.39, 2.0; standard deviation = 0.874.
  - Resulting human capital distribution mean and standard deviation: 0.565 and 0.108.
  - Occupational advantage ε = [√0.5 + ˜ε, √0.5 − ˜ε] with ˜ε ∼ Truncated Normal(0, σ_ε; 0.5); σ_ε calibrated to target volatility.

### Exogenous and endogenous parameter values (selected)
- Table 5 key parameter values (as calibrated for the UK benchmark economy):
  - T_r = 0.000 (Baseline)
  - z_R = 1.000 (Normalization)
  - r = 0.010 (Long-run UK government bonds rate)
  - α = 2.000 (Conservative intertemporal elasticity of substitution)
  - β = 0.960 (Gross interest rates inverse)
  - γ = 0.700 (Shimer (2005))
  - ν = 1.600 (Schaal (2012))
  - ρ_R = 0.422 (Aggregate routine AR(1) process)
  - σ_R = 0.015 (Aggregate routine AR(1) process)
  - ρ_N = 0.366 (Aggregate non-routine AR(1) process)
  - σ_N = 0.008 (Aggregate non-routine AR(1) process)
  - z_N = 0.955 (Routine employment share calibration)
  - T_u = 0.155 (Unemployment benefits transfers-GDP ratio)
  - δ = 0.010 (Conservative exogenous separation rate)
  - χ = 0.001 (Unemployment rate during GFC)
  - α_Bh = 7.255 (Labor income mean distribution)
  - β_Bh = 4.855 (Labor income dispersion distribution)
  - σ_ε = 0.355 (Labor income volatility)
- Benchmark retention policy transfers set to zero pre-GFC (retention transfers almost negligible in the U.K. prior to the global financial crisis).

### Calibration targets and model fit (selected moments)
- Calibration targets include: pre-GFC and GFC unemployment rates, routine-employment share, unemployment-benefits-to-GDP ratio, cross-sectional distribution and volatility of normalized labor income.
- Table 6: Data vs Model statistics targets
  - Routine employment share: Data 0.553, Model 0.547
  - Unemployment benefits-GDP ratio: Data 0.002, Model 0.002
  - Unemployment rate before GFC: Data 0.051, Model 0.052
  - Unemployment rate during GFC: Data 0.072, Model 0.069
  - Labor income normalized percentiles (data): 0.01/0.04/0.09/0.15/0.22
  - Labor income normalized percentiles (model): 0.08/0.14/0.21/0.27/0.34
  - Volatility normalized labor income: Data 0.874, Model 0.779
- Calibration notes:
  - Model matches unemployment and routine share closely.
  - Earnings moments are flatter in the model (two occupations + Beta-distributed human capital) leading simulated volatility 0.779 vs empirical 0.874.
  - Authors note allowing more within-occupation wage heterogeneity would close earnings gaps without altering policy conclusions, which hinge on:
    - (i) relative earnings loss after separation,
    - (ii) employment dynamics.

_Italic: Source: wpiea2026015-source-pdf - 1.  Given the labor market tightness_

### 6.2    Quantitative Results

### wpiea2026015-source-pdf - 6.2    Quantitative Results

### Quantitative framework and experiment design
- Shock design and timing:
  - An asymmetric negative shock in productivity affects both occupations, replicating the U.K. experience during the global financial crisis and the European Debt Crisis when unemployment hit record-high levels in 2011.
  - Routine occupations experienced a deeper contraction in employment than non-routine occupations, at 6.9 and 2.4 percentage points, respectively.
  - The shock lasts for one period, after which productivity returns to average levels.
  - The transition back to the steady state lasts for 10 periods overall.
  - The transition consists of 10 periods, representing 2 years and 6 months.
- Fiscal mechanics during the shock:
  - Negative aggregate shock increases government spending on unemployment insurance and reduces labor income tax collections.
  - To satisfy the government inter-temporal budget constraint, government issues public debt financed with a non-distortionary fixed lump-sum tax during each period of the transition.
  - Labor income tax rates are kept fixed.

### Baseline economy (no retention policy)
- Labor market dynamics:
  - After the shock, the unemployment rate increased by 1.6 percentage points.
  - The share of workers in routine occupations fell by almost two percentage points.
  - Employment in non-routine occupations increased by less than one percentage point.
  - Worker reallocation: workers with high levels of human capital in routine and non-routine occupations move to non-routine occupations because of the asymmetric nature of the shock.
- Mechanism:
  - A negative productivity shock diminishes firms' productivity; with fixed wages and no borrowing, unprofitable firm-worker matches dissolve, raising unemployment.

### Retention policy design and labor-market effects
- Policy objective and mechanism:
  - The retention policy aims to retain half of the otherwise lost matches after the negative aggregate shock by setting the maximum retention transfer.
  - The policy subsidizes wages of workers in matches that become unprofitable, making firms indifferent between firing and retaining workers.
- Labor-market outcomes with retention transfers:
  - Unemployment rises by 0.8 percentage points with the retention policy (instead of 1.6 percentage points under the baseline).
  - Employment in routine occupations falls by 1.5 percentage points with the policy instead of 2 percentage points in the baseline.
  - Employment in non-routine occupations rises by 0.8 percentage points with the policy instead of 0.5 percentage points in the baseline.
- Summary of retention policy effects:
  - Reduces the spike in the unemployment rate.
  - Speeds up the transition back to the steady state because fewer workers end up unemployed.
  - Reduces occupation switches from routine to non-routine occupations.
  - These three channels reduce the economic cost of the recession.

### Fiscal costs and budget dynamics
- Key drivers of fiscal costs:
  - i) costs associated with unemployment benefits; ii) labor tax revenues; iii) transfers provided by the retention policy scheme.
- Government spending and revenue outcomes (model with retention scheme relative to baseline):
  - Unemployment insurance spending decreases from 2.7% of GDP in the baseline economy to 2.4% in the economy with retention policy schemes, resulting in a decrease of 0.3% in government spending.
  - The higher employment in the retention-policy economy increases government revenue from labor income tax compared to the economy without the policy.
  - The retention policy comes with a small fiscal cost of almost 0.02% of GDP compared to the 0.3% savings in unemployment insurance.
- Fiscal deficit dynamics:
  - Baseline model: the fiscal deficit is 0.6 percentage points during the period of the shock, and it takes about seven periods for the baseline model to return to a balanced budget.
  - With the retention policy: the fiscal deficit is 0.3 percentage points during the period of the shock (half of the baseline impact despite retention policy cost), and it takes five periods to return to a balanced budget.
  - Interpretation: the retention policy halves the fiscal deficit during the shock period and reduces the time to return to a balanced budget.

### Welfare analysis and distributional impacts
- Distribution of retention transfers:
  - The three lowest human capital deciles accrue about 50% of all retention transfers.
  - The routine occupation accrues 61% of all the transfers.
  - Matches with lower human capital are less profitable and more vulnerable to negative profits after the aggregate shock; matches with high human capital are less susceptible.
- Aggregate welfare impact:
  - Implementing a retention policy yields positive welfare gains of about a 0.06% increase in permanent consumption-equivalent.
- Heterogeneous welfare gains across human capital:
  - Lower deciles benefit the most; the lowest deciles’ welfare increases by almost 0.15%.
  - Higher human capital workers benefit less because they are less vulnerable to the negative aggregate shock.
  - Net effect: all workers are better off with the retention policy because the government fiscal deficit is lower, implying lower transfer payments for workers.

### Robustness: Longer transitory shocks (two-period shock)
- Shock design:
  - The negative productivity shock lasts for two periods (periods 0 and 1) and then returns to steady state over the following nine periods.
  - Three scenarios considered: no retention policy, retention policy only in the first shock period, retention policy in both shock periods.
- Unemployment dynamics (two-period shock):
  - No retention transfers: unemployment rises from 7.3 percent in period 0 to 8.2 percent in period 1.
  - Retention transfers only in the first period: unemployment reduced initially to 6.4 percent in period 0, but rises to 7.7 percent in period 1 after the policy is removed.
  - Retention transfers in both shock periods: unemployment limited to 6.3 percent in period 0 and 6.8 percent in period 1, stabilizing employment across both periods.
  - Interpretation: when the shock is more persistent, covering the full duration of the shock is essential to avoid further job losses and secure a faster recovery.
- Fiscal deficit dynamics (two-period shock, relative to economy without retention transfers):
  - Retention transfers only in the first period: deficit declines by 0.31 percent of GDP in period 0 but improvement weakens to a 0.19 percent reduction in period 1 once policy is removed.
  - Retention transfers in both shock periods: the fiscal position improves substantially and persistently; the deficit drops by 0.36 percent in period 0 and continues to deepen to a

*Source: wpiea2026015-source-pdf - 6.2    Quantitative Results*

### 0.53 percent reduction in period 1, illustrating that preventing job separations throughout the

### wpiea2026015-source-pdf - 0.53 percent reduction in period 1, illustrating that preventing job separations throughout the

### Effects on fiscal burden and unemployment during transitory shocks
- A fully implemented retention policy further eases the fiscal burden by preventing job separations throughout the full duration of a negative shock, with the fiscal deviation gradually narrowing as employment normalizes during the transition to steady state.
- A 0.53 percent reduction in period 1 exemplifies the fiscal improvement when separations are prevented at the onset of the shock.
- Retention support stabilizes labor market conditions more effectively and reduces government crisis-related spending needs during negative shocks.

### Welfare gains from retention support (two-period transitory shock)
- Welfare gains measured as permanent consumption-equivalent changes relative to no policy:
  - Retention transfers implemented only in the first period: average welfare gain ≈ 0.07 percent.
  - Providing support throughout both shock periods: average welfare gain ≈ 0.14 percent.
- The doubling of lifetime welfare benefits from single-period to full-duration coverage is consistent across all deciles of human capital.
- Partial implementation becomes slightly more beneficial when downturns are more persistent:
  - One-period shock with one-period policy: average welfare gain ≈ 0.06 percent.
  - Two-period shock with policy applied only in the first period: average welfare gain ≈ 0.07 percent.
- Overall: extending support to each period of a temporary downturn materially strengthens lifetime welfare gains.

### Permanent shock comparison (2.4 percentage points permanent productivity decline)
- Shock design:
  - A permanent decline in productivity of 2.4 percentage points applied symmetrically to both occupations (homogeneous shock to isolate persistence).
- Unemployment dynamics:
  - Without retention transfers, unemployment adjusts almost immediately to new steady state level of 6 percent.
  - One-period retention transfers temporarily reduce unemployment by preventing initial separations, but as support is withdrawn and fundamentals remain permanently weaker, unemployment rises and converges to the same steady state level of 6 percent.
  - Conclusion: one-period retention primarily delays separations rather than preventing long-run match destruction under a permanent shock.
- Fiscal effects:
  - Implementing retention transfers for one period delivers only a modest fiscal improvement, "on the order of one fifth of a percentage point of GDP."
  - Limited fiscal benefit reflects preservation of matches that are not viable under permanently weaker productivity; policy mainly postpones separations rather than generating durable expansion in taxable labor income.
- Welfare effects:
  - Average welfare gain under the permanent shock is positive but modest, at around 0.03%.
  - Mechanism: temporary retention keeps workers attached to employment enabling on-the-job search and supporting bargaining power, reducing welfare costs of joblessness even when preserved matches are not viable long-run.

### Retention policy costs sensitivity (bureaucratic cost added)
- Assumption change: add bureaucratic cost per unit transferred, interpreted as transfer fees, processing salaries, private information search costs, or implementation inefficiencies.
- Cost calibration during crisis period:
  - Full bureaucratic cost: 1% of GDP.
  - Half bureaucratic cost: 0.5% of GDP.
- Implication given calibration:
  - For every 1£ spent on retention transfers, the government spends 31£ or 63£ on bureaucratic costs, respectively (when bureaucratic costs amount to 0.5% and 1% of GDP as calibrated).
  - Baseline policy cost (no bureaucratic cost) is 0.02% of GDP.
- Financing:
  - The bureaucratic cost is financed via non-distortionary lump-sum transfers adjusted to satisfy the government intertemporal budget constraint.
- Fiscal and welfare impacts:
  - Increasing bureaucratic cost by 0.5 or 1 percentage point of GDP similarly increases the government deficit after the shock (Panel 10(a) dynamics).
  - Half a percent of GDP bureaucratic cost "has almost the same fiscal cost as in the baseline period, exceeding it by 2 percentage points."
  - Adding bureaucratic costs increases lump-sum transfers imposed on all workers, reducing consumption and lowering welfare for every agent.
  - With a substantial bureaucratic cost of 1% of GDP, the retention policy ceases to be a Pareto improvement:
    - Workers with higher levels of human capital can experience welfare losses because their matches are profitable even under the negative shock, while the bureaucratic cost offsets potential policy savings.

### Key quantitative results and robustness
- Model calibration and main quantitative outcomes:
  - A retention scheme calibrated to offset roughly half of the observed rise in unemployment would have meaningfully dampened labor-market deterioration in the calibrated historical episodes.
  - Welfare rises by 0.06 percent in consumption-equivalent terms in the baseline transitory-shock experiment.
  - Welfare gains are concentrated among low-income workers who benefit most from match preservation and avoidance of costly reallocation.
  - The policy can be fiscally cost-effective: by reducing unemployment-insurance outlays and preserving the tax base, the scheme can lower the fiscal deficit despite the wage subsidy.
- Robustness:
  - Welfare gains increase in longer downturns.
  - The policy remains welfare-improving even when the adverse shock is permanent, with retention facilitating adjustment via on-the-job search rather than unemployment.

_Italic source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026015-source-pdf.pdf_

### References

### Annex A (contained after References)

### A.1 Country list in the sample (Table A.1)
- Table A.1: Country List (columns: Country Name, Country Code, Year, Avg Annual Obs, Total Obs)
  - 1 Austria AT 2004-2018 7,168 107,516
  - 2 Belgium BE 2004-2018 7,151 107,270
  - 3 Denmark DK 2003-2018 5,096 81,542
  - 4 Estonia EE 2005-2018 6,157 86,201
  - 5 Finland FI 2004-2018 11,239 168,589
  - 6 Italy IT 2004-2018 23,942 359,132
  - 7 Netherlands NL 2005-2018 12,414 173,791
  - 8 Portugal PT 2004-2017 6,859 96,020
  - 9 Slovenia SI 2005-2018 15,203 212,847
  - 10 Spain ES 2004-2018 17,346 260,189
  - 11 Sweden SE 2004-2018 7,331 109,969
  - 12 United Kingdom UK 2005-2016 9,733 116,795
  - Total 129,639 1,879,861

### A.2 Mathematical proofs — main propositions and lemmas (statements and conclusions)
- A.2.1 Proof of Lemma 1
  - Starting expression:
    - ∂∆k/∂φk = ∆k(φk,h;Tr)(1−γ)−φk > 0.
  - Conclusion: "the measure of job matches saved increases the deeper the recession."
  - Additional expression:
    - ∂∆k/∂h = −∆k(φk,h;Tr) / h < 0
  - Conclusion: "the measure of job matches saved decrease as the human capital increases."

- A.2.2 Proof of Proposition 1
  - Proved three statements individually:
    - i) For two recessions φ2 and φ1 with φk2 ≥ φk1 > 0 and φk′2 = φk′1 for all k ≠ k′ and φk2, φk1 ≤ (1−γ)(1− Tu z k H0) for every occupation k. Using Lemma 1 and Assumption??:
      - f(h)(εk1(h,φk2) − εkr(h,φk2)) > f(h)(εk1(h,φk1) − εkr(h,φk1))
      - Aggregating:
        - ∑_{k=1}^K ∫_{H0}^∞ (εk1(h,φk2) − εkr(h,φk2)) dF(h) >
          ∑_{k=1}^K ∫_{H0}^∞ (εk1(h,φk1) − εkr(h,φk1)) dF(h).
      - Conclusion: "under a deeper recession a retention policy scheme saves more matches."
    - ii) For a recession φ > 0 with φk ≤ (1−γ)(1− Tu zk H0) for every occupation:
      - f(h)(εk1(h,φk) − εkr(h,φk)) = f(h) Tr((1−γ)−φk) z h.
      - H(H) ≡ ∑_{k=1}^K ∫_{H0}^H (εk1(h,φk) − εkr(h,φk)) dF(h) is strictly decreasing.
      - Conclusion: "more jobs are saved in matches with a lower level of human-capital."
    - iii) For an asymmetric recession φ > 0 with φ1 ≤ φ2 ≤ ... ≤ φK ≤ (1−γ)(1− Tu z H0):
      - εk1(h,φk) − εkr(h,φk) = Tr((1−γ)−φk) z h.
      - For k < k′ and φk < φk′ :
        - ∫_{H0}^∞ Tr((1−γ)−φk) z h dF(h) < ∫_{H0}^∞ Tr((1−γ)−φk′) z h dF(h).
      - Conclusion: "a retention policy scheme saves more firms’ matches in occupations affected the most under an asymmetric recession."

- A.2.3 Proof of Proposition 2
  - Savings expression for occupation k and human capital h:
    - ∫_{εkr}^{εk1} (Tu + Ik(ε,h,(1−φ)z)) dε = (Tr^2 − Tu) Tr((1−γ)−φk) z k h ≥ 0, because Tr ≤ 2 Tu.
  - Aggregating across h and k:
    - Tr^2 (Tr − 2 Tu) ∑_{k=1}^K 1/((1−γ)−φk) (∫_{H0}^∞ 1/h dF(h)) ≥ 0.
  - Conclusion: "a retention policy scheme saves more low-productive firms’ matches."

### A.3 Employment by occupation — UK (Figure 11)
- Figure 11: UK Employment by Occupation
  - Notes:
    - The figure plots the employment in routine and non-routine occupation in the UK (years 2005–2016 plotted on x-axis).
    - Source: Office of National Statistics (ONS).
  - Visual range (as shown):
    - y-axis plotted from 0.30 to 0.60 (employment share).
    - Time series labeled 2005 through 2016.
  - Legend:
    - Employment Routine
    - Employment Non-routine

### A.4 Baseline versus counterfactual analysis for government series (Figure 12)
- Figure 12: Government budget decomposition and transition baseline versus counterfactual
  - Notes:
    - x-axis represents the periods of the model. The solid blue line represents the baseline series without retention policy scheme and the red dashed line the counterfactual with retention policy scheme.
    - Assumption: the negative productive shock happens in period 0 and takes ten periods to return to the steady state. Periods before the shock represent the stationary equilibrium.
    - y-axis represents the expenditures and revenues of the fiscal budget as a percentage of GDP in the model with retention policy scheme relative to the baseline model.
  - Panels shown:
    - (a) Fiscal deficit (y-axis labeled 0.0% to 1.5%)
    - (b) Unemployment benefits expenditure (y-axis labeled 2.0% to 3.5%)
    - (c) Labor tax revenues (y-axis labels showing 2.207% to 2.211%)
    - (d) Retention transfers expenditure (y-axis from -0.01% to 0.05%)

*Source: wpiea2026015-source-pdf - References (PDF chapter/section content provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026015-source-pdf.pdf_